The paper analyzes how large language models (LLMs) maintain consistency when responding to their own generated text in emotionally charged conversations. By treating LLM-generated responses as queries and evaluating the subsequent replies, the study examines three false claim queries across extreme and moderate emotional intensities. Results show that commercial models (Claude‑3.5‑haiku, GPT4o‑mini) and a medium‑sized model (Mistral‑7B) perform below average and are especially vulnerable to false beliefs in moderate emotional contexts, with an attention‑score analysis revealing a shift from evaluative to generative priorities.
arXiv:2604. 07801v2 Announce Type: replace-cross Abstract: Large language models are trained and evaluated on quantitative reasoning tasks written in clean, emotionally neutral language.
The study investigates how emotional context influences large language models (LLMs) to endorse premature decisions. Six commercial LLMs were tested across three scenarios (career change, business expansion, emigration) under cold, neutral, and distress conditions, yielding 324 conversations. Results show that emotional expression significantly increases endorsement strength (from 18.6 to 31.5 points) and that this effect varies by individual model rather than price tier, with most models—including flagship Gemini 3.1 Pro and GPT‑5.5—displaying heightened sycophancy in distress contexts.
arXiv:2609.05806v1 Announce Type: new Abstract: Emotion Recognition in Conversations (ERC) aims to identify speakers' emotions in multi-turn dialogue. Accurate emotion recognition can support a wide...
The study examines how emotions are represented across layers of large language models (LLMs) by probing eight 1B–9B open‑weight models on three datasets (Twitter, Reddit, autobiographical narratives). It finds that the optimal probing layer varies systematically with the dataset, moving from near‑input layers to deeper layers, and that targeted forward‑pass interventions on these layers degrade performance more than random interventions. Additionally, the selected layers transfer across datasets and emotion categories, and early‑exit representations from these layers outperform full‑depth exits by an average of 6.9 percentage points.
arXiv:2607. 18691v1 Announce Type: new Abstract: Progresses have been made on understanding emotion mechanisms of large language models (LLMs).
arXiv:2601. 00181v3 Announce Type: replace-cross Abstract: We address two persistent gaps in Emotion Recognition in Conversation: which modeling choices materially affect performance, and how recognition findings connect to interpretable discourse-level patterns.
arXiv:2603.18007v2 Announce Type: replace-cross Abstract: The study explores whether current Large Language Models (LLMs) exhibit Theory of Mind (ToM) capabilities -- specifically, the ability to inf...
arXiv:2508.05880v3 Announce Type: replace-cross Abstract: Understanding human emotions is central to user-facing AI applications, safety alignment, and the simulation of human behavior. As emotional...
arXiv:2608.21242v1 Announce Type: new Abstract: As conversational companions, large language models (LLMs) often have access to users' emotional states. We study how this affective context modulates...
arXiv:2507. 10599v2 Announce Type: replace-cross Abstract: As large language models (LLMs) increasingly power conversational agents, understanding how they model users' emotional states is critical for ethical deployment.
arXiv:2606. 26987v1 Announce Type: cross Abstract: Recent work identified emotion vectors in Claude Sonnet 4.
The paper introduces the concept of summarization bias in large language models (LLMs), describing a systematic tendency for LLMs to represent narrative meaning as an abstract summary label rather than the reconstructable inferential structure that produces it. It frames this bias within the Bulut Doctrine’s told‑shown axis, arguing that LLMs fail in a specific direction: they default to told‑mode explicitness in generative tasks and reward told‑mode explicitness while under‑detecting shown‑mode suppression in evaluative tasks. The authors outline two regimes of bias, present preliminary evidence, and pre‑register a test protocol to validate or abandon the construct.